γ₀: General Multi-Embodiment Robot Control via Reinforcement Learning